For nearly a quarter-century, researchers have struggled to understand how apolipoprotein E modifies risk for Alzheimer’s disease. A new study offers a beguilingly simple explanation: ApoE elevates the transcription of amyloid precursor protein (APP), and each isoform does so a manner that parallels the AD risk it confers. Reporting in Cell on January 26, researchers led by Thomas Südhof at Stanford University describe a signaling pathway in which ApoE engages a receptor on human neurons, stabilizes a protein kinase, and ultimately promotes the transcription of APP and production of Aβ. ApoE4 switched on this pathway more robustly than ApoE3, which did so more than ApoE2. While ApoE4’s dominance over other isoforms in promoting APP expression was modest, Südhof told Alzforum that over a lifetime, it could add up to a large cumulative risk in E4 carriers.

While researchers praised the results, some doubted a single mechanism could entirely account for ApoE-related AD risk. “In complex conditions such as AD, which involve effects over many years of aging, single pathways will not be enough to explain subtle changes to pathogenesis over time,” commented William Rebeck of Georgetown University in Washington, D.C. 

Signals of Susceptibility?

With a potency that parallels their genetic risk profile, ApoE isoforms kick off a signaling cascade that enhances transcription of APP and production of Aβ. [Image courtesy of Huang et al., Cell 2017.]

Many questions remain about this signaling cascade, including the identity of the ApoE receptor, how ApoE isoforms activate it differentially, and where this cascade fits among myriad other ApoE functions in the human brain that could alter AD risk, either dependently or independently of Aβ.

ApoE functions as a transporter of lipids in the brain, where it latches onto various receptors to facilitate internalization of its lipid cargo. ApoE4 carriers have an increased risk of developing AD, while ApoE2 carriers enjoy some protection. Efforts to understand ApoE’s relationship with AD have produced many possible explanations, some controversial. For one thing, Aβ and ApoE interact, and AD mice expressing the human ApoE4 gene have more Aβ plaques, soluble Aβ oligomers, and cognitive problems than animals expressing human ApoE2. Slower clearance of Aβ is widely considered a leading explanation of at least some of ApoE4’s effect (see Jul 2011 webinar), and blocking Aβ/ApoE interactions ameliorates amyloid pathology (see Pankiewicz et al., 2014). Researchers have reported that halving ApoE expression in mice or treating them with antibodies directed against ApoE facilitates plaque removal, as might bexarotene, which increases lipidation of ApoE (see Dec 2011 newsKim et al., 2012; May 2014 news). Furthermore, soluble Aβ and ApoE compete for binding and internalization by the lipoprotein receptor 1 (LPR1) in an isoform-dependent manner, suggesting that ApoE4 may prevent Aβ clearance by reducing its uptake via competition (see Apr 2013 news). ApoE4 also appears to harm neurons independently of Aβ (see Feb 2012 conference news). 

Against this dizzying backdrop of findings, mostly in mice, first author Yu-Wen Alvin Huang and colleagues asked a simple question: What proteins secreted from glial cells affect Aβ production in human neurons? In an effort to isolate and examine these factors, the researchers had to grow neurons separately from glia. They coaxed human excitatory neurons from embryonic stem cells, and maintained them on a supportive layer of mouse embryonic fibroblasts (MEFs), rather than the standard mouse glia. Compared to human neurons grown on glia, those grown on MEFs produced less than half as much Aβ40 and Aβ42. Treating the MEF-supported neurons with 24 of the most abundantly expressed proteins from mouse glia, the researchers found three—ApoE, Igf2, and IgfBP2—that significantly increased Aβ production. They decided to continue investigating the relationship between ApoE and APP expression.

Moving beyond the single isoform of mouse ApoE, the researchers found that MEF-supported human neurons produced more APP mRNA, as well as Aβ40, Aβ42, and total Aβ, when treated with recombinant human ApoE4 than when given ApoE3 or ApoE2. APP mRNA induction ranged from three- to fivefold, while increases in Aβ were more modest. Aβ42 concentrations increased from approximately 20 pg/mL in untreated neurons to 25, 30, and 35 pg/mL for ApoE2-, ApoE3-, and ApoE4-treated neurons, respectively. The same was true for two different human iPSC-derived neuron lines grown on MEFs. However, treating glia-supported neurons with human ApoE of any isoform did not additionally boost APP or Aβ production, presumably because glial-derived ApoE or other glial proteins maxed out the signal. Strikingly, ApoE did not boost the expression of two close APP homologues, APLP or APLP2.

The researchers next unspooled the signaling cascade linking ApoE and APP. First of all, they found that ApoE triggered phosphorylation of the MAP kinase ERK1/2, and that inhibiting ApoE binding to surface receptors blocked this phosphorylation, as well as APP and Aβ. Using inhibition, knock-down, and overexpression approaches to fill in the missing parts of the pathway, the researchers found that ApoE engaged cell surface receptors that stimulated the protection of the MAP kinase kinase kinase DLK from destruction by the proteasome. DLK then phosphorylated the MAP kinase kinase MKK7, which then did the same to ERK1/2. This last piece of the puzzle was surprising, as previous studies had indicated that DLK primarily triggers the phosphorylation of Jun kinases (JNKs), not ERK1/2. That the different ApoE isoforms stimulated each step of the pathway with the same potency ranking, namely ApoE4>ApoE3>ApoE2, was satisfying, said Südhof.

To hunt down the transcription factor responsible for upping APP expression, the researchers employed CRISPR interference (aka CRISPRi). They used a series of guide RNAs, along with the catalytically dead form of Cas9 (dCas9), to target and block transcription initiation from different regions within the APP promoter. They found that one guide RNA, corresponding to the AP-1 transcription factor binding site, blocked ApoE’s stimulation of APP transcription. Dimeric AP-1 transcription factors consist of a combination of cFos, cJun, or other proteins. Accordingly, the researchers found that ApoE (again, with a potency rank of ApoE4>ApoE3>ApoE2) stimulated the phosphorylation of cFos, and that expressing a dominant negative version of cFos in human neurons wiped out ApoE’s induction of APP.

The researchers next extended their findings into mature mouse neurons and live mice. Though simply adding ApoE to neuron/glial cultures did not boost neuronal APP expression (likely because other glial-derived signals had maxed out expression), knocking down DLK reduced the phosphorylation of MKK7 and ERK1/2, as well as levels of APP mRNA, protein, and Aβ. In what the researchers claim is the first time CRISPRi has been attempted in vivo, they injected viruses expressing dCas9 and AP-1 binding site directed guide RNAs into the cortices of newborn mice. They detected a reduction in APP mRNA, indicating that the mouse APP promoter, like its human counterpart, is regulated by AP-1. Injecting viruses expressing the dominant negative version of cFos had the same effect. Due to the multitude of glial factors secreted in the mouse brain, the researchers did not directly test whether ApoE turned on this pathway there.

Südhof proposed that the differential activation of this AP-1 signaling pathway by ApoE isoforms could account for the differences in AD risk among genotypes. However, he was careful to point out that how APP expression and Aβ production subsequently affect AD risk was unclear. He told Alzforum that he hopes to uncover the ApoE receptor that initiates the cascade, as well as any membrane-associated molecules, such as scaffolding proteins, that veer the pathway toward DLK stabilization.

David Holtzman of Washington University in St. Louis found the experiments that uncovered the ApoE-triggered DLK signaling cascade convincing. “It’s beautiful work. At this point, it is not clear whether the effect of ApoE on increasing APP and Aβ is relevant in the brain of animals or humans,” he told Alzforum. “That’s the next major question that needs to be answered.” Holtzman pointed out that the physiological form of ApoE secreted by astrocytes in the brain is more highly lipidated than the ApoE that Huang and colleagues used in their study, which were either recombinant forms or secreted by HEK293 cells. Furthermore, ApoE may also change its properties as it associates with Aβ aggregates in the AD brain, he added. How these and other in vivo complexities affect the pathway is unclear, Holtzman said. “It is clear from a lot of prior work that ApoE affects Aβ clearance and aggregation. If it also affects Aβ synthesis in the mammalian brain, then this mechanism that [Huang and colleagues] describe will be key to further work,” he said. “It could help us develop ways to lower Aβ.”

Guojun Bu of the Mayo Clinic in Jacksonville, Florida, echoed Holtzman’s comment about the lipidation state of ApoE. “ApoE/lipoprotein particles secreted by astrocytes are truly lipidated particles and should be tested in these assays,” he wrote to Alzforum. He added that the relevance of the pathway could be tested further in ApoE targeted replacement mice, which express different human ApoE isoforms.

Patrick Sullivan of Duke University in Durham, North Carolina, who first generated the ApoE targeted replacement mice, added that future studies could utilize astrocytic cell lines that he and Holtzman generated from those mice as a source of more physiologically lipidated human ApoE. Still, Sullivan said that Südhof and colleagues’ thorough, reductionist approach allowed them to home in on a novel pathway with new potential targets. He speculated that ApoE4’s heightened ability to trigger the pathway may stem from its increased affinity for the as-yet-unidentified receptor, in keeping with the isoform’s established superior affinity for other known receptors in the LDL receptor family.

Sullivan added that the predominant hypothesis used to explain ApoE-related AD risk revolves around the protein’s effects on Aβ clearance, while others think inflammation plays a key role. “This work brings in an entirely new mechanism that may in part explain the increased amyloid burden in E4 carriers,” he said. He added that this regulatory pathway could also get researchers closer to addressing a fundamental question that has nagged the field for decades: What is the physiological function of APP and Aβ in the brain?—Jessica Shugart


  1. The work of Huang et al. is very interesting in its finding that one of the mechanisms of ApoE altering Alzheimer’s disease risk is through neuronal signal transduction, involving DLK, ERK, and cFos. The data show that the three ApoE isoforms differentially affect APP transcription and Aβ production, with effects that parallel ApoE-related risk (ApoE2<ApoE3<ApoE4). An interesting aspect of this model is that ApoE is affecting brain function before the appearance of pathological changes, and thus dependent on the authors using models of normal tissue. Interestingly, the glial factors initially identified that induced APP and Aβ were related to changes in lipid metabolism (ApoE lipoproteins) and glucose metabolism (insulin growth factor 2 and its binding protein); these factors may help explain the strong acute induction of APP after many models of brain injury.

    Technically, this work shows the great promise of CRISPRi technology in defining numerous members of complex pathways. In complex conditions like AD, which involve effects over many years of aging, single pathways will not be enough to explain subtle changes to pathogenesis over time. Future work in this area will surely examine whether there is evidence of the ApoE effects on this signal transduction pathway in control brains in mice and humans, adding data on chronic effects to the current data on acute effects.

  2. Previous studies have supported differential effects of ApoE isoforms on Aβ clearance, aggregation and toxicity. This study, using human neurons, suggests that ApoE isoform effects relevant to AD might also include impact on Aβ production through modulating APP expression. Specific contribution of individual pathways to AD risk require further studies using in vivo models and in humans.

    The cellular and biochemical experiments elegantly showed that ApoE isoforms have different effects (ApoE4>ApoE3>ApoE2) on stimulating a signaling pathway in neurons that includes the DLK/MKK7/ERK1/2 MAP kinase pathway and cFos phosphorylation, impacting APP gene transcription. It will be interesting to examine whether such signaling has physiological functions by perhaps influencing neuronal and synaptic functions.

    Validation using in vivo model systems expressing different human APOE gene alleles will be important. For example, the effects of ApoE gene alleles on APP expression could be evaluated using human APOE targeted replacement (TR) mice. Changes in expression of the endogenous mouse APP gene or the human APP gene (e.g., YAC-APP Tg mice) as a result of ApoE isoform expression (targeted replacement or AAV viral mediated) could be evaluated.  

    A critical relevance of this work is that it was performed using ES- and iPS cell-derived human neurons. However, the relevance to human brain, in particular aging brains that are more relevant to AD, requires further investigation.

    The authors indicated that ApoE particles secreted by HEK293 cells are lipidated, containing cholesterol, however, there is little support for such a conclusion from the literature or this study. In fact, published work suggests that ApoE particles secreted by HEK cells are not lipidated, rather they are mostly just ApoE aggregates (LaDu et al., 2006). ApoE/lipoprotein particles secreted by astrocytes (primary cells from ApoE-TR mice or iPSC-derived) are truly lipidated particles and should be tested in these assays.

    The results using the lipoprotein receptor antagonist RAP are interesting, as they suggest an involvement of RAP-responsive ApoE receptors, which could either be ApoE signaling receptors (e.g., ApoER2 and VLDLR) or metabolic receptors (e.g., LRP1 and LDLR). Further studies are needed to sort out the receptor(s) involved. That could help elucidate the physiological and pathological pathways relevant to ApoE and/or the receptor.


    . Self-assembly of HEK cell-secreted ApoE particles resembles ApoE enrichment of lipoproteins as a ligand for the LDL receptor-related protein. Biochemistry. 2006 Jan 17;45(2):381-90. PubMed.

Make a Comment

To make a comment you must login or register.


Webinar Citations

  1. Slow Aβ; Clearance Is ApoE4’s Modus Operandi in Late-Onset AD

News Citations

  1. Lowering ApoE Brings Down Amyloid in Mice
  2. Has ApoE’s Time Come as a Therapeutic Target?
  3. ApoE Does Not Bind Aβ, Competes for Clearance
  4. San Francisco: GABA Neurons Blamed for Memory Loss in ApoE Mice

Paper Citations

  1. . Blocking the apoE/Aβ interaction ameliorates Aβ-related pathology in APOE ε2 and ε4 targeted replacement Alzheimer model mice. Acta Neuropathol Commun. 2014 Jun 28;2:75. PubMed.
  2. . Anti-apoE immunotherapy inhibits amyloid accumulation in a transgenic mouse model of Aβ amyloidosis. J Exp Med. 2012 Nov 19;209(12):2149-56. PubMed.

Other Citations

  1. ApoE targeted replacement mice

Further Reading

Primary Papers

  1. . ApoE2, ApoE3, and ApoE4 Differentially Stimulate APP Transcription and Aβ Secretion. Cell. 2017 Jan 26;168(3):427-441.e21. Epub 2017 Jan 19 PubMed.